Lung Abnormality Detection using Deep learning
Principal Investigator
Name
Harshil Soni
Degrees
B.E.
Institution
System Level Solutions India Ptv.Ltd.
Position Title
Project Manager
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-875
Initial CDAS Request Approval
Jan 26, 2022
Title
Lung Abnormality Detection using Deep learning
Summary
Machine learning-based models can be used as assistant clinicians in managing incidental or screen-detected indeterminate pulmonary nodules. Such systems may be able to reduce variability in nodule classification, improve decision making and ultimately reduce the number of benign nodules that are needlessly followed or worked up.
We aim to develop an automatic lung abnormality detection system that detects abnormal parts of the lung from the CT scan images and also classifies them.
The NLST database provides access to several CT scans that are expository to the development of a deep learning-based model. The NLST data would be used for training the discriminative models, along with an expressive sample of lung abnormalities. Our Team of ML engineers will use all available imaging and clinical NLST data to train predictive models that will be used to detect and classify the abnormal parts. The Team will also consult radiologists for regular guidance as well as outcome validation.
Our final goal is to develop such type of system that upgrades the diagnostic process and clinical task.
We aim to develop an automatic lung abnormality detection system that detects abnormal parts of the lung from the CT scan images and also classifies them.
The NLST database provides access to several CT scans that are expository to the development of a deep learning-based model. The NLST data would be used for training the discriminative models, along with an expressive sample of lung abnormalities. Our Team of ML engineers will use all available imaging and clinical NLST data to train predictive models that will be used to detect and classify the abnormal parts. The Team will also consult radiologists for regular guidance as well as outcome validation.
Our final goal is to develop such type of system that upgrades the diagnostic process and clinical task.
Aims
1. Develop deep learning models for abnormality detection, segmentation, and classification.
2. Develop an interactive system that provides locations of abnormal parts, their size and intensity, probability of abnormality, and malignancy classification. It will also allow user interactions to criticize and amend the software results.
3. The developed system will be used as an assistant to Radiologists.
Collaborators
Mikita Parikh, System Level Solutions India Ptv.Ltd.
Tejas Vaghela, System Level Solutions India Ptv.Ltd.